基于Gabor小波和颜色矩的棉花盲椿象分类方法

Classification of cotton blind stinkbug based on Gabor wavelet and color moments

为了提高棉花盲椿象测报精度,提出了一种在自然环境下基于Gabor小波和颜色矩的棉花盲椿象自动分类方法。该方法以5种棉花盲椿象为研究对象,利用Gabor小波和颜色矩分别提取盲椿象图像的纹理和颜色特征,并结合主成分分析和支持向量机,实现了棉花盲椿象的分类。通过特征提取与优化试验发现,利用主成分变换得到的第1主成分、第2主成分和第3主成分分量累计贡献率为87.3%,且聚类效果较好;通过棉花盲椿象分类试验得出,经过主成分分析与径向基核函数支持向量机相结合的棉花盲椿象分类效果最好,其训练时间为89 ms,分类正确率达91%。该方法能准确对棉花盲椿象进行分类与识别。

The method of classifying cotton blind stinkbugs based on Gabor wavelet and color moments was developed for improving the forecasting and warning of blind stinkbug disaster under field conditions.In this method,Gabor wavelet and color moments were used to extract texture features and color features from cotton blind stinkbugs,in order to detect cotton blind stinkbug with principal component analysis and support vector machine.The results of feature extraction and optimization showed that cumulative contribution rate of first three principal components was 87.3%,and first three principal components obtained better clustering results.Cotton blind stinkbug classification results found combination of first three principal components and radial basis function support vector machine better than others.Training time and accuracy rate of the methods were 89 ms and 91%,respectively.The study results showed that the proposed classification method could be used to accurately classify the five kind of blind stinkbugs.